wip: initial simd support

Signed-off-by: Henry Gressmann <mail@henrygressmann.de>

10 files changed, 150 insertions, 12 deletions

diff --git a/crates/cli/src/args.rs b/crates/cli/src/args.rs
index 96c3605..a23f1f4 100644
--- a/crates/cli/src/args.rs
+++ b/crates/cli/src/args.rs
@@ -25,6 +25,7 @@ impl FromStr for WasmArg {
             "i64" => val.parse::<i64>().map_err(|e| format!("invalid argument value for i64: {e:?}"))?.into(),
             "f32" => val.parse::<f32>().map_err(|e| format!("invalid argument value for f32: {e:?}"))?.into(),
             "f64" => val.parse::<f64>().map_err(|e| format!("invalid argument value for f64: {e:?}"))?.into(),
+            "v128" => val.parse::<u128>().map_err(|e| format!("invalid argument value for v128: {e:?}"))?.into(),
             t => return Err(format!("Invalid arg type: {t}")),
         };
 
diff --git a/crates/tinywasm/Cargo.toml b/crates/tinywasm/Cargo.toml
index 2cad2e4..0ea406c 100644
--- a/crates/tinywasm/Cargo.toml
+++ b/crates/tinywasm/Cargo.toml
@@ -32,11 +32,12 @@ serde_json={version="1.0"}
 serde={version="1.0", features=["derive"]}
 
 [features]
-default=["std", "parser", "logging", "archive"]
+default=["std", "parser", "logging", "archive", "simd"]
 logging=["log", "tinywasm-parser?/logging", "tinywasm-types/logging"]
 std=["tinywasm-parser?/std", "tinywasm-types/std"]
 parser=["dep:tinywasm-parser"]
 archive=["tinywasm-types/archive"]
+simd=[]
 
 [[test]]
 name="test-wasm-1"
diff --git a/crates/tinywasm/src/interpreter/executor.rs b/crates/tinywasm/src/interpreter/executor.rs
index 3c579c1..291ab64 100644
--- a/crates/tinywasm/src/interpreter/executor.rs
+++ b/crates/tinywasm/src/interpreter/executor.rs
@@ -3,10 +3,15 @@
 use super::no_std_floats::NoStdFloatExt;
 
 use alloc::{format, rc::Rc, string::ToString};
-use core::ops::ControlFlow;
+use core::ops::{ControlFlow, Not};
+use core::simd::cmp::{SimdPartialEq, SimdPartialOrd};
+use core::simd::num::SimdUint;
 use interpreter::stack::CallFrame;
 use tinywasm_types::*;
 
+#[cfg(feature = "simd")]
+use super::simd::*;
+
 use super::num_helpers::*;
 use super::stack::{BlockFrame, BlockType, Stack};
 use super::values::*;
@@ -41,6 +46,8 @@ impl<'store, 'stack> Executor<'store, 'stack> {
     #[inline(always)]
     fn exec_next(&mut self) -> ControlFlow<Option<Error>> {
         use tinywasm_types::Instruction::*;
+
+        #[rustfmt::skip]
         match self.cf.fetch_instr() {
             Nop | BrLabel(_) | I32ReinterpretF32 | I64ReinterpretF64 | F32ReinterpretI32 | F64ReinterpretI64 => {}
             Unreachable => self.exec_unreachable()?,
@@ -302,6 +309,75 @@ impl<'store, 'stack> Executor<'store, 'stack> {
             LocalCopy128(from, to) => self.exec_local_copy::<Value128>(*from, *to),
             LocalCopyRef(from, to) => self.exec_local_copy::<ValueRef>(*from, *to),
 
+            V128Not => self.stack.values.replace_top_same::<Value128>(|v| Ok(!v)).to_cf()?,
+            V128And => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a & b)).to_cf()?,
+            V128AndNot => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a & (!b))).to_cf()?,
+            V128Or => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a | b)).to_cf()?,
+            V128Xor => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a ^ b)).to_cf()?,
+            V128Bitselect => self.stack.values.calculate_same_3::<Value128>(|v1, v2, c| Ok((v1 & c) | (v2 & !c))).to_cf()?,
+            V128AnyTrue => self.stack.values.replace_top::<Value128, i32>(|v| Ok((v.reduce_sum() != 0) as i32)).to_cf()?,
+
+            I8x16Swizzle => self.stack.values.calculate_same::<Value128>(|a, s| Ok(a.swizzle_dyn(s))).to_cf()?,
+
+            I8x16Splat => self.stack.values.replace_top::<i32, Value128>(|v| Ok(Simd::<i8, 16>::splat(v as i8).to_ne_bytes())).to_cf()?,
+            I16x8Splat => self.stack.values.replace_top::<i32, Value128>(|v| Ok(Simd::<i16, 8>::splat(v as i16).to_ne_bytes())).to_cf()?,
+            I32x4Splat => self.stack.values.replace_top::<i32, Value128>(|v| Ok(Simd::<i32, 4>::splat(v).to_ne_bytes())).to_cf()?,
+            I64x2Splat => self.stack.values.replace_top::<i64, Value128>(|v| Ok(Simd::<i64, 2>::splat(v).to_ne_bytes())).to_cf()?,
+            F32x4Splat => self.stack.values.replace_top::<f32, Value128>(|v| Ok(Simd::<f32, 4>::splat(v).to_ne_bytes())).to_cf()?,
+            F64x2Splat => self.stack.values.replace_top::<f64, Value128>(|v| Ok(Simd::<f64, 2>::splat(v).to_ne_bytes())).to_cf()?,
+
+            I8x16Eq => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_eq(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8Eq => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_eq(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4Eq => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_eq(b).to_int().to_ne_bytes())).to_cf()?,
+            F32x4Eq => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_eq(b).to_int().to_ne_bytes())).to_cf()?,
+            F64x2Eq => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_eq(b).to_int().to_ne_bytes())).to_cf()?,
+
+            I8x16Ne => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ne(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8Ne => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ne(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4Ne => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ne(b).to_int().to_ne_bytes())).to_cf()?,
+            F32x4Ne => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ne(b).to_int().to_ne_bytes())).to_cf()?,
+            F64x2Ne => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ne(b).to_int().to_ne_bytes())).to_cf()?,
+
+            I8x16LtS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8LtS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4LtS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+            I64x2LtS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+            F32x4Lt => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+            F64x2Lt => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+
+            I8x16LtU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8LtU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4LtU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_lt(b).to_int().to_ne_bytes())).to_cf()?,
+            I64x2GtS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+            F32x4Gt => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+            F64x2Gt => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+
+            I8x16GtS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8GtS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4GtS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+            I64x2LeS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+            F32x4Le => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+            F64x2Le => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+
+            I8x16GtU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8GtU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4GtU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_gt(b).to_int().to_ne_bytes())).to_cf()?,
+            I64x2GeS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ge(b).to_int().to_ne_bytes())).to_cf()?,
+            F32x4Ge => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ge(b).to_int().to_ne_bytes())).to_cf()?,
+            F64x2Ge => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ge(b).to_int().to_ne_bytes())).to_cf()?,
+
+            I8x16LeS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8LeS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4LeS => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+
+            I8x16LeU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8LeU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4LeU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_le(b).to_int().to_ne_bytes())).to_cf()?,
+
+            I8x16GeU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ge(b).to_int().to_ne_bytes())).to_cf()?,
+            I16x8GeU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ge(b).to_int().to_ne_bytes())).to_cf()?,
+            I32x4GeU => self.stack.values.calculate_same::<Value128>(|a, b| Ok(a.simd_ge(b).to_int().to_ne_bytes())).to_cf()?,
+
             i => return ControlFlow::Break(Some(Error::UnsupportedFeature(format!("unimplemented opcode: {i:?}")))),
         };
 
diff --git a/crates/tinywasm/src/interpreter/mod.rs b/crates/tinywasm/src/interpreter/mod.rs
index 0b7df2f..269bcfb 100644
--- a/crates/tinywasm/src/interpreter/mod.rs
+++ b/crates/tinywasm/src/interpreter/mod.rs
@@ -6,6 +6,9 @@ mod values;
 #[cfg(not(feature = "std"))]
 mod no_std_floats;
 
+#[cfg(feature = "simd")]
+mod simd;
+
 use crate::{Result, Store};
 pub use values::*;
 
diff --git a/crates/tinywasm/src/interpreter/simd.rs b/crates/tinywasm/src/interpreter/simd.rs
new file mode 100644
index 0000000..53c64e3
--- /dev/null
+++ b/crates/tinywasm/src/interpreter/simd.rs
@@ -0,0 +1,18 @@
+pub(super) use core::ops::Neg;
+
+pub(super) use core::simd::Simd;
+pub(super) use core::simd::ToBytes;
+pub(super) use core::simd::num::SimdFloat;
+pub(super) use core::simd::num::SimdInt;
+
+macro_rules! impl_wasm_simd_val {
+    ($($v:ident),*) => {
+        $(
+            pub(super) fn $v(f: core::simd::u8x16) -> core::simd::$v {
+                core::simd::$v::from_ne_bytes(f)
+            }
+        )*
+    };
+}
+
+impl_wasm_simd_val!(i8x16, i16x8, i32x4, i64x2, f32x4, f64x2);
diff --git a/crates/tinywasm/src/interpreter/stack/value_stack.rs b/crates/tinywasm/src/interpreter/stack/value_stack.rs
index 9830c01..843723a 100644
--- a/crates/tinywasm/src/interpreter/stack/value_stack.rs
+++ b/crates/tinywasm/src/interpreter/stack/value_stack.rs
@@ -72,6 +72,11 @@ impl ValueStack {
     }
 
     #[inline]
+    pub(crate) fn calculate_same_3<T: InternalValue>(&mut self, func: impl FnOnce(T, T, T) -> Result<T>) -> Result<()> {
+        T::stack_calculate3(self, func)
+    }
+
+    #[inline]
     pub(crate) fn calculate<T: InternalValue, U: InternalValue>(
         &mut self,
         func: impl FnOnce(T, T) -> Result<U>,
@@ -170,11 +175,16 @@ impl ValueStack {
         match val_type {
             ValType::I32 => WasmValue::I32(self.pop()),
             ValType::I64 => WasmValue::I64(self.pop()),
-            ValType::V128 => WasmValue::V128(self.pop()),
             ValType::F32 => WasmValue::F32(self.pop()),
             ValType::F64 => WasmValue::F64(self.pop()),
             ValType::RefExtern => WasmValue::RefExtern(ExternRef::new(self.pop())),
             ValType::RefFunc => WasmValue::RefFunc(FuncRef::new(self.pop())),
+
+            #[cfg(not(feature = "simd"))]
+            ValType::V128 => WasmValue::V128(self.pop()),
+
+            #[cfg(feature = "simd")]
+            ValType::V128 => WasmValue::V128(u128::from_ne_bytes(self.pop::<Value128>().to_array())),
         }
     }
 
diff --git a/crates/tinywasm/src/interpreter/values.rs b/crates/tinywasm/src/interpreter/values.rs
index 712baf7..7a2a916 100644
--- a/crates/tinywasm/src/interpreter/values.rs
+++ b/crates/tinywasm/src/interpreter/values.rs
@@ -5,9 +5,13 @@ use super::stack::{Locals, ValueStack};
 
 pub(crate) type Value32 = u32;
 pub(crate) type Value64 = u64;
-pub(crate) type Value128 = u128;
 pub(crate) type ValueRef = Option<u32>;
 
+#[cfg(feature = "simd")]
+pub(crate) type Value128 = core::simd::u8x16;
+#[cfg(not(feature = "simd"))]
+pub(crate) type Value128 = u128;
+
 #[derive(Debug, Clone, Copy, PartialEq)]
 /// A untyped WebAssembly value
 pub enum TinyWasmValue {
@@ -106,9 +110,14 @@ impl TinyWasmValue {
             ValType::I64 => WasmValue::I64(self.unwrap_64() as i64),
             ValType::F32 => WasmValue::F32(f32::from_bits(self.unwrap_32())),
             ValType::F64 => WasmValue::F64(f64::from_bits(self.unwrap_64())),
-            ValType::V128 => WasmValue::V128(self.unwrap_128()),
             ValType::RefExtern => WasmValue::RefExtern(ExternRef::new(self.unwrap_ref())),
             ValType::RefFunc => WasmValue::RefFunc(FuncRef::new(self.unwrap_ref())),
+
+            #[cfg(feature = "simd")]
+            ValType::V128 => WasmValue::V128(u128::from_ne_bytes(self.unwrap_128().to_array())),
+
+            #[cfg(not(feature = "simd"))]
+            ValType::V128 => WasmValue::V128(self.unwrap_128()),
         }
     }
 }
@@ -118,11 +127,16 @@ impl From<&WasmValue> for TinyWasmValue {
         match value {
             WasmValue::I32(v) => TinyWasmValue::Value32(*v as u32),
             WasmValue::I64(v) => TinyWasmValue::Value64(*v as u64),
-            WasmValue::V128(v) => TinyWasmValue::Value128(*v),
             WasmValue::F32(v) => TinyWasmValue::Value32(v.to_bits()),
             WasmValue::F64(v) => TinyWasmValue::Value64(v.to_bits()),
             WasmValue::RefExtern(v) => TinyWasmValue::ValueRef(v.addr()),
             WasmValue::RefFunc(v) => TinyWasmValue::ValueRef(v.addr()),
+
+            #[cfg(not(feature = "simd"))]
+            WasmValue::V128(v) => TinyWasmValue::Value128(*v),
+
+            #[cfg(feature = "simd")]
+            WasmValue::V128(v) => TinyWasmValue::Value128(v.to_ne_bytes().into()),
         }
     }
 }
@@ -146,6 +160,9 @@ pub(crate) trait InternalValue: sealed::Sealed + Into<TinyWasmValue> {
     fn stack_calculate(stack: &mut ValueStack, func: impl FnOnce(Self, Self) -> Result<Self>) -> Result<()>
     where
         Self: Sized;
+    fn stack_calculate3(stack: &mut ValueStack, func: impl FnOnce(Self, Self, Self) -> Result<Self>) -> Result<()>
+    where
+        Self: Sized;
 
     fn stack_pop(stack: &mut ValueStack) -> Self
     where
@@ -203,6 +220,19 @@ macro_rules! impl_internalvalue {
                 }
 
                 #[inline(always)]
+                fn stack_calculate3(stack: &mut ValueStack, func: impl FnOnce(Self, Self, Self) -> Result<Self>) -> Result<()> {
+                    let v3 = stack.$stack.pop();
+                    let v2 = stack.$stack.pop();
+                    let v1 = stack.$stack.last_mut();
+                    let (Some(v1), Some(v2), Some(v3)) = (v1, v2, v3) else {
+                         unreachable!("ValueStack underflow, this is a bug");
+                    };
+
+                    *v1 = $to_internal(func($to_outer(*v1), $to_outer(v2), $to_outer(v3))?);
+                    return Ok(())
+                }
+
+                #[inline(always)]
                 fn replace_top(stack: &mut ValueStack, func: impl FnOnce(Self) -> Result<Self>) -> Result<()> {
                     let Some(v) = stack.$stack.last_mut() else {
                         unreachable!("ValueStack underflow, this is a bug");
diff --git a/crates/tinywasm/src/lib.rs b/crates/tinywasm/src/lib.rs
index d3431e2..72c38cd 100644
--- a/crates/tinywasm/src/lib.rs
+++ b/crates/tinywasm/src/lib.rs
@@ -5,7 +5,7 @@
 ))]
 #![warn(missing_docs, missing_debug_implementations, rust_2018_idioms, unreachable_pub)]
 #![forbid(unsafe_code)]
-// #![cfg_attr(feature = "nightly", feature(portable_simd))]
+#![cfg_attr(feature = "simd", feature(portable_simd))]
 
 //! A tiny WebAssembly Runtime written in Rust
 //!
diff --git a/crates/tinywasm/tests/generated/wasm-3.csv b/crates/tinywasm/tests/generated/wasm-3.csv
index ed301e4..26f2937 100644
--- a/crates/tinywasm/tests/generated/wasm-3.csv
+++ b/crates/tinywasm/tests/generated/wasm-3.csv
@@ -1 +1 @@
-0.9.0-alpha.0,32305,2591,[{"name":"address.wast","passed":260,"failed":0},{"name":"address64.wast","passed":0,"failed":242},{"name":"align.wast","passed":161,"failed":0},{"name":"align64.wast","passed":83,"failed":73},{"name":"annotations.wast","passed":74,"failed":0},{"name":"binary-leb128.wast","passed":92,"failed":1},{"name":"binary.wast","passed":126,"failed":0},{"name":"block.wast","passed":223,"failed":0},{"name":"br.wast","passed":97,"failed":0},{"name":"br_if.wast","passed":119,"failed":0},{"name":"br_on_non_null.wast","passed":1,"failed":9},{"name":"br_on_null.wast","passed":1,"failed":9},{"name":"br_table.wast","passed":24,"failed":162},{"name":"bulk.wast","passed":117,"failed":0},{"name":"call.wast","passed":91,"failed":0},{"name":"call_indirect.wast","passed":49,"failed":124},{"name":"call_ref.wast","passed":4,"failed":31},{"name":"comments.wast","passed":8,"failed":0},{"name":"const.wast","passed":778,"failed":0},{"name":"conversions.wast","passed":619,"failed":0},{"name":"custom.wast","passed":11,"failed":0},{"name":"data.wast","passed":59,"failed":6},{"name":"elem.wast","passed":137,"failed":14},{"name":"endianness.wast","passed":69,"failed":0},{"name":"endianness64.wast","passed":0,"failed":69},{"name":"exports.wast","passed":97,"failed":0},{"name":"f32.wast","passed":2514,"failed":0},{"name":"f32_bitwise.wast","passed":364,"failed":0},{"name":"f32_cmp.wast","passed":2407,"failed":0},{"name":"f64.wast","passed":2514,"failed":0},{"name":"f64_bitwise.wast","passed":364,"failed":0},{"name":"f64_cmp.wast","passed":2407,"failed":0},{"name":"fac.wast","passed":8,"failed":0},{"name":"float_exprs.wast","passed":927,"failed":0},{"name":"float_literals.wast","passed":179,"failed":0},{"name":"float_memory.wast","passed":90,"failed":0},{"name":"float_memory64.wast","passed":0,"failed":90},{"name":"float_misc.wast","passed":471,"failed":0},{"name":"forward.wast","passed":5,"failed":0},{"name":"func.wast","passed":175,"failed":0},{"name":"func_ptrs.wast","passed":36,"failed":0},{"name":"global.wast","passed":53,"failed":71},{"name":"i32.wast","passed":460,"failed":0},{"name":"i64.wast","passed":416,"failed":0},{"name":"id.wast","passed":7,"failed":0},{"name":"if.wast","passed":241,"failed":0},{"name":"imports.wast","passed":169,"failed":90},{"name":"inline-module.wast","passed":1,"failed":0},{"name":"instance.wast","passed":0,"failed":23},{"name":"int_exprs.wast","passed":108,"failed":0},{"name":"int_literals.wast","passed":51,"failed":0},{"name":"labels.wast","passed":29,"failed":0},{"name":"left-to-right.wast","passed":96,"failed":0},{"name":"linking.wast","passed":142,"failed":21},{"name":"load.wast","passed":118,"failed":0},{"name":"load64.wast","passed":59,"failed":38},{"name":"local_get.wast","passed":36,"failed":0},{"name":"local_init.wast","passed":10,"failed":0},{"name":"local_set.wast","passed":53,"failed":0},{"name":"local_tee.wast","passed":98,"failed":0},{"name":"loop.wast","passed":120,"failed":0},{"name":"memory-multi.wast","passed":6,"failed":0},{"name":"memory.wast","passed":89,"failed":0},{"name":"memory64.wast","passed":14,"failed":53},{"name":"memory_copy.wast","passed":8385,"failed":515},{"name":"memory_fill.wast","passed":164,"failed":36},{"name":"memory_grow.wast","passed":157,"failed":0},{"name":"memory_grow64.wast","passed":0,"failed":49},{"name":"memory_init.wast","passed":307,"failed":173},{"name":"memory_redundancy.wast","passed":8,"failed":0},{"name":"memory_redundancy64.wast","passed":0,"failed":8},{"name":"memory_size.wast","passed":49,"failed":0},{"name":"memory_trap.wast","passed":182,"failed":0},{"name":"memory_trap64.wast","passed":0,"failed":172},{"name":"names.wast","passed":486,"failed":0},{"name":"nop.wast","passed":88,"failed":0},{"name":"obsolete-keywords.wast","passed":11,"failed":0},{"name":"ref.wast","passed":12,"failed":1},{"name":"ref_as_non_null.wast","passed":1,"failed":6},{"name":"ref_func.wast","passed":17,"failed":0},{"name":"ref_is_null.wast","passed":2,"failed":20},{"name":"ref_null.wast","passed":0,"failed":34},{"name":"return.wast","passed":84,"failed":0},{"name":"return_call.wast","passed":11,"failed":34},{"name":"return_call_indirect.wast","passed":26,"failed":50},{"name":"return_call_ref.wast","passed":11,"failed":40},{"name":"select.wast","passed":155,"failed":2},{"name":"skip-stack-guard-page.wast","passed":11,"failed":0},{"name":"stack.wast","passed":7,"failed":0},{"name":"start.wast","passed":20,"failed":0},{"name":"store.wast","passed":111,"failed":0},{"name":"switch.wast","passed":28,"failed":0},{"name":"table-sub.wast","passed":2,"failed":1},{"name":"table.wast","passed":47,"failed":9},{"name":"table_copy.wast","passed":1742,"failed":30},{"name":"table_copy_mixed.wast","passed":4,"failed":0},{"name":"table_fill.wast","passed":45,"failed":35},{"name":"table_get.wast","passed":16,"failed":1},{"name":"table_grow.wast","passed":58,"failed":21},{"name":"table_init.wast","passed":780,"failed":96},{"name":"table_set.wast","passed":26,"failed":2},{"name":"table_size.wast","passed":39,"failed":1},{"name":"tag.wast","passed":1,"failed":8},{"name":"throw.wast","passed":3,"failed":10},{"name":"throw_ref.wast","passed":2,"failed":13},{"name":"token.wast","passed":61,"failed":0},{"name":"traps.wast","passed":36,"failed":0},{"name":"try_table.wast","passed":11,"failed":51},{"name":"type-canon.wast","passed":0,"failed":2},{"name":"type-equivalence.wast","passed":12,"failed":20},{"name":"type-rec.wast","passed":6,"failed":14},{"name":"type.wast","passed":3,"failed":0},{"name":"unreachable.wast","passed":64,"failed":0},{"name":"unreached-invalid.wast","passed":121,"failed":0},{"name":"unreached-valid.wast","passed":2,"failed":11},{"name":"unwind.wast","passed":50,"failed":0},{"name":"utf8-custom-section-id.wast","passed":176,"failed":0},{"name":"utf8-import-field.wast","passed":176,"failed":0},{"name":"utf8-import-module.wast","passed":176,"failed":0},{"name":"utf8-invalid-encoding.wast","passed":176,"failed":0}]
+0.9.0-alpha.0,32305,2598,[{"name":"address.wast","passed":260,"failed":0},{"name":"address64.wast","passed":0,"failed":242},{"name":"align.wast","passed":161,"failed":0},{"name":"align64.wast","passed":83,"failed":73},{"name":"annotations.wast","passed":74,"failed":0},{"name":"binary-leb128.wast","passed":92,"failed":1},{"name":"binary.wast","passed":126,"failed":0},{"name":"block.wast","passed":223,"failed":0},{"name":"br.wast","passed":97,"failed":0},{"name":"br_if.wast","passed":119,"failed":0},{"name":"br_on_non_null.wast","passed":1,"failed":9},{"name":"br_on_null.wast","passed":1,"failed":9},{"name":"br_table.wast","passed":24,"failed":162},{"name":"bulk.wast","passed":117,"failed":0},{"name":"call.wast","passed":91,"failed":0},{"name":"call_indirect.wast","passed":49,"failed":124},{"name":"call_ref.wast","passed":4,"failed":31},{"name":"comments.wast","passed":8,"failed":0},{"name":"const.wast","passed":778,"failed":0},{"name":"conversions.wast","passed":619,"failed":0},{"name":"custom.wast","passed":11,"failed":0},{"name":"data.wast","passed":59,"failed":6},{"name":"elem.wast","passed":137,"failed":14},{"name":"endianness.wast","passed":69,"failed":0},{"name":"endianness64.wast","passed":0,"failed":69},{"name":"exports.wast","passed":97,"failed":0},{"name":"f32.wast","passed":2514,"failed":0},{"name":"f32_bitwise.wast","passed":364,"failed":0},{"name":"f32_cmp.wast","passed":2407,"failed":0},{"name":"f64.wast","passed":2514,"failed":0},{"name":"f64_bitwise.wast","passed":364,"failed":0},{"name":"f64_cmp.wast","passed":2407,"failed":0},{"name":"fac.wast","passed":8,"failed":0},{"name":"float_exprs.wast","passed":927,"failed":0},{"name":"float_literals.wast","passed":179,"failed":0},{"name":"float_memory.wast","passed":90,"failed":0},{"name":"float_memory64.wast","passed":0,"failed":90},{"name":"float_misc.wast","passed":471,"failed":0},{"name":"forward.wast","passed":5,"failed":0},{"name":"func.wast","passed":175,"failed":0},{"name":"func_ptrs.wast","passed":36,"failed":0},{"name":"global.wast","passed":53,"failed":71},{"name":"i32.wast","passed":460,"failed":0},{"name":"i64.wast","passed":416,"failed":0},{"name":"id.wast","passed":7,"failed":0},{"name":"if.wast","passed":241,"failed":0},{"name":"imports.wast","passed":169,"failed":90},{"name":"inline-module.wast","passed":1,"failed":0},{"name":"instance.wast","passed":0,"failed":23},{"name":"int_exprs.wast","passed":108,"failed":0},{"name":"int_literals.wast","passed":51,"failed":0},{"name":"labels.wast","passed":29,"failed":0},{"name":"left-to-right.wast","passed":96,"failed":0},{"name":"linking.wast","passed":142,"failed":21},{"name":"load.wast","passed":118,"failed":0},{"name":"load64.wast","passed":59,"failed":38},{"name":"local_get.wast","passed":36,"failed":0},{"name":"local_init.wast","passed":10,"failed":0},{"name":"local_set.wast","passed":53,"failed":0},{"name":"local_tee.wast","passed":98,"failed":0},{"name":"loop.wast","passed":120,"failed":0},{"name":"memory-multi.wast","passed":6,"failed":0},{"name":"memory.wast","passed":89,"failed":1},{"name":"memory64.wast","passed":14,"failed":55},{"name":"memory_copy.wast","passed":8385,"failed":515},{"name":"memory_fill.wast","passed":164,"failed":36},{"name":"memory_grow.wast","passed":157,"failed":0},{"name":"memory_grow64.wast","passed":0,"failed":49},{"name":"memory_init.wast","passed":307,"failed":173},{"name":"memory_redundancy.wast","passed":8,"failed":0},{"name":"memory_redundancy64.wast","passed":0,"failed":8},{"name":"memory_size.wast","passed":49,"failed":0},{"name":"memory_trap.wast","passed":182,"failed":0},{"name":"memory_trap64.wast","passed":0,"failed":172},{"name":"names.wast","passed":486,"failed":0},{"name":"nop.wast","passed":88,"failed":0},{"name":"obsolete-keywords.wast","passed":11,"failed":0},{"name":"ref.wast","passed":12,"failed":1},{"name":"ref_as_non_null.wast","passed":1,"failed":6},{"name":"ref_func.wast","passed":17,"failed":0},{"name":"ref_is_null.wast","passed":2,"failed":20},{"name":"ref_null.wast","passed":0,"failed":34},{"name":"return.wast","passed":84,"failed":0},{"name":"return_call.wast","passed":11,"failed":34},{"name":"return_call_indirect.wast","passed":26,"failed":50},{"name":"return_call_ref.wast","passed":11,"failed":40},{"name":"select.wast","passed":155,"failed":2},{"name":"skip-stack-guard-page.wast","passed":11,"failed":0},{"name":"stack.wast","passed":7,"failed":0},{"name":"start.wast","passed":20,"failed":0},{"name":"store.wast","passed":111,"failed":0},{"name":"switch.wast","passed":28,"failed":0},{"name":"table-sub.wast","passed":2,"failed":1},{"name":"table.wast","passed":47,"failed":13},{"name":"table_copy.wast","passed":1742,"failed":30},{"name":"table_copy_mixed.wast","passed":4,"failed":0},{"name":"table_fill.wast","passed":45,"failed":35},{"name":"table_get.wast","passed":16,"failed":1},{"name":"table_grow.wast","passed":58,"failed":21},{"name":"table_init.wast","passed":780,"failed":96},{"name":"table_set.wast","passed":26,"failed":2},{"name":"table_size.wast","passed":39,"failed":1},{"name":"tag.wast","passed":1,"failed":8},{"name":"throw.wast","passed":3,"failed":10},{"name":"throw_ref.wast","passed":2,"failed":13},{"name":"token.wast","passed":61,"failed":0},{"name":"traps.wast","passed":36,"failed":0},{"name":"try_table.wast","passed":11,"failed":51},{"name":"type-canon.wast","passed":0,"failed":2},{"name":"type-equivalence.wast","passed":12,"failed":20},{"name":"type-rec.wast","passed":6,"failed":14},{"name":"type.wast","passed":3,"failed":0},{"name":"unreachable.wast","passed":64,"failed":0},{"name":"unreached-invalid.wast","passed":121,"failed":0},{"name":"unreached-valid.wast","passed":2,"failed":11},{"name":"unwind.wast","passed":50,"failed":0},{"name":"utf8-custom-section-id.wast","passed":176,"failed":0},{"name":"utf8-import-field.wast","passed":176,"failed":0},{"name":"utf8-import-module.wast","passed":176,"failed":0},{"name":"utf8-invalid-encoding.wast","passed":176,"failed":0}]
diff --git a/crates/types/src/instructions.rs b/crates/types/src/instructions.rs
index c61a02f..e3e30f6 100644
--- a/crates/types/src/instructions.rs
+++ b/crates/types/src/instructions.rs
@@ -206,12 +206,11 @@ pub enum Instruction {
     F32x4ExtractLane(u8), F32x4ReplaceLane(u8),
     F64x2ExtractLane(u8), F64x2ReplaceLane(u8),
 
-    V128Not, V128And, V128AndNot, V128Or, V128Xor, V128Bitselect, V128AnyTrue,
-
-    I8x16Splat, I8x16Swizzle, I8x16Eq, I8x16Ne, I8x16LtS, I8x16LtU, I8x16GtS, I8x16GtU, I8x16LeS, I8x16LeU, I8x16GeS, I8x16GeU,
+    V128Not, V128And, V128AndNot, V128Or, V128Xor, V128Bitselect, V128AnyTrue, I8x16Swizzle,
+    I8x16Splat, I8x16Eq, I8x16Ne, I8x16LtS, I8x16LtU, I8x16GtS, I8x16GtU, I8x16LeS, I8x16LeU, I8x16GeS, I8x16GeU,
     I16x8Splat, I16x8Eq, I16x8Ne, I16x8LtS, I16x8LtU, I16x8GtS, I16x8GtU, I16x8LeS, I16x8LeU, I16x8GeS, I16x8GeU,
     I32x4Splat, I32x4Eq, I32x4Ne, I32x4LtS, I32x4LtU, I32x4GtS, I32x4GtU, I32x4LeS, I32x4LeU, I32x4GeS, I32x4GeU,
-    I64x2Splat, I64x2Eq, I64x2Ne, I64x2LtS, I64x2GtS, I64x2LeS, I64x2GeS, 
+    I64x2Splat, I64x2Eq, I64x2Ne, I64x2LtS, I64x2GtS, I64x2LeS, I64x2GeS,
     F32x4Splat, F32x4Eq, F32x4Ne, F32x4Lt, F32x4Gt, F32x4Le, F32x4Ge,
     F64x2Splat, F64x2Eq, F64x2Ne, F64x2Lt, F64x2Gt, F64x2Le, F64x2Ge,